eve Agent with Python Template

eve Agent with Python Template

A starter template for building a web chat application where an eve AI agent manages a shared task list through a Python FastAPI backend, displayed on a Next.js frontend and deployable on Vercel Services.

What is the eve Agent with Python Template?

The eve Agent with Python Template is an open-source starter that combines an eve agent, a Next.js web chat frontend, and a Python FastAPI backend into a single Vercel Services project. Visitors chat with the agent, which calls FastAPI endpoints to create, update, and delete tasks; the same tasks appear on the app's Tasks page. Vercel Labs publishes the template under the MIT License.

Key Features

• Shared task list — the agent calls the same FastAPI endpoints as the Tasks page via list_tasks, create_task, and update_task tools, so chat and page stay in sync.
• Python compute tools — compute_stats returns mean, median, and quartiles, while analyze_text returns word counts, reading time, readability, and common words, all calculated in Python.
• Generative UI with json-render — the render_ui tool produces result cards, metrics, and tables from a fixed component catalog.
• Approval-gated deletion — the delete_task tool asks for user approval before removing a task.
• Password access control — visitors sign in with the password set in the EVE_CHAT_PASSWORD environment variable; the password can be replaced with Clerk or Better Auth.
• Backend visibility pages — a Backend page shows the running Python and FastAPI versions plus a statistics form, and an API page lists every endpoint from the OpenAPI document.
• Next.js, FastAPI, and Neon Postgres — the frontend uses Next.js with TanStack Table; the backend uses FastAPI with SQLModel and stores tasks in Neon Postgres (or a local SQLite file during development).
• Local dev tooling — Node.js 24, pnpm, and uv are used; pnpm dev:all runs the web app, agent, and backend together with the Vercel CLI.

Who is it for?

• AI developers who want a working reference for connecting an eve agent to a Python backend and a Next.js chat interface rather than wiring the stack from scratch.
• Python backend developers who need to expose FastAPI endpoints as agent tools and see how tool calls map to shared application state.
• Teams prototyping internal tools that combine conversational task management with Python-based text or statistics analysis.

What can you do with the eve Agent with Python Template?

• Task management via chat: ask the agent to add, complete, or delete tasks; changes appear on the Tasks page and persist in Postgres.
• Text analysis: paste text into the chat and get word count, reading time, readability scores, and the most frequent words computed by the Python backend.
• Statistics: request mean, median, and quartiles for a dataset, calculated through the same endpoint exposed on the Backend page.
• Custom agent UI: extend the component catalog in agent/lib/ui-catalog.ts and apps/web/app/_components/ui-registry.tsx to let the agent render new card types.

How does the eve Agent with Python Template work?

The template deploys the agent, web app, and FastAPI backend together through Vercel Services; vercel.ts attaches the agent with withEve and passes the backend URL in BACKEND_URL. The backend has no public route, so browsers reach it only through the web app, and the agent calls its endpoints as tools.

FAQ

Is the eve Agent with Python Template free?

Yes. The repository is published under the MIT License and can be cloned and deployed from GitHub. Running it on Vercel requires Vercel and Neon accounts for the project, database, and AI Gateway model access.

What is required to get started?

You need an EVE_CHAT_PASSWORD environment variable for sign-in and a Neon Postgres database, which sets DATABASE_URL for the backend. The Vercel clone flow creates the repository and project and prompts for both.

Can I use my own authentication instead of the password?

Yes. The customization guide describes replacing EVE_CHAT_PASSWORD with Clerk, Better Auth, or another auth provider. With the built-in password, everyone who has it shares one task list and can see each other's chats and tasks.

What Python and frontend technologies does it use?

The template uses Next.js with TanStack Table for the frontend, FastAPI with SQLModel for the backend, Neon Postgres for storage, and uv for Python package management; json-render and Vercel Sandbox support generative UI and sandboxed execution.